Charts – doing it properly
Updated: Nov 23, 2020
I cannot speak for all industries as a whole but in Finance when presenting a report or results of a study or anything really you need to make sure that: a) your data is correct and findings are relevant b) the presentation is clear and as good looking as possible. Of course, having the right results is the most important thing, but the final success of your presentation is truly shared 50/50 between the data and the way you chose to display it. I need to stress again, that when talking about presentation, I mean both the visuals and the clarity – it doesn’t matter if you have the best looking charts if no one can understand their underlying meaning.
Simple, simple, simple
Try to do things as simple as possible. Always show your results in the most commonly used chart types, mainly because you are not doing the presentation for yourself, but rather for someone else (your approach should be tailored to them). Chances are they are not doing data analysis on a daily basis and are used to reading simpler charts. On the other hand, if the data is too complicated to display simply, maybe you have not done such a great job analysing and aggregating it.
For everything else, adopt the less is more principle.
Don’t try to do too much in one chart. There is no point of showing new sales, portfolio outstanding ad bad loans volume all in the same chart, even if you can plot them on one axis. Those would be better off with 3 separate charts.
Hide the labels when creating a pivot chart.
There is no point in having both labels and a data table. If you need data labels – keep them only on the important parts and not everywhere.
If you have only one series, remove the legend.
Axis titles are generally needed only if you have more than one vertical axis.
Your grid lines should not overpower the main data. 50% transparency and grey colour usually do miracles for this.
Along with some general principles:
Avoid “fancy” chart types such as radar and area charts.
Don’t use logarithmic scales unless you really have to.
Don’t try to be fancy with error bars and trend lines, unless that is the main point. If you do have them, first make sure you yourself understand what they mean. Especially if you are plotting the R-squared and the equation on a date axis, can you explain why the X coefficient is so small?
Your title should not state the obvious. If data in your chart is really self-explanatory you do not need a title at all.
Always try to maximise the area of the chart itself. Not the clutter.
Format your numbers and dates for readability. Date axis labels do not need to show days if you are aggregating months. Rotate them for easier reading. Truncate very large numbers but always give the reader an indication for this.
Even the most commonly used (line) chart, can look like an overdecorated Christmas tree, even if there is only one series in it.
Simple is always better
Do you really need a chart?
Yes, it’s great that you finally found data which you can represent in a surface chart or a 3D scatter plot but do you really need to do it? Isn’t there a simpler way to show the data?
Let’s assume we want to examine concentrations of risk grades among the “loan purpose”. As both the purpose and the grade have some prior distributions, we need to calculate this as a pivot table index.
Let’s see how we can plot the results.
We can try a surface plot. The purposes with a high index are visible, but the grades are too many to plot properly. Also, the low index values are hard to identify. Different rotations won’t help a lot and anyone reading might not get the point.
When plotted as a radar, we can see all the grades. The logarithmic scale helps a lot to make the low index visible, so we can identify both the top and the bottom performers. Still, the chart is confusing and the readability is not great – we cannot see clearly the difference between the separate “loan purposes”.
Why not just show directly the pivot table! It is compact enough, so it takes no more space than a chart. All the categories are easily visible and readable. The magnitude of the difference in the index is both visible in colour and with the numbers directly. The conditional formatting adds a visual representation and we could technically call this a “heat map chart”.
Rule of 6
A general rule of thumb which I use for the basic charts is (number of series) x (number of axes) <= 6. The axes, in this case, are the primary/secondary ones. This leaves two possible combinations: primary axis + 6 data series, primary + secondary axis + 3 data series (total, not each). I usually extend this rule when building dashboards, to a maximum number of chars per screen <= 6 and a maximum number of slicers <= 6. But overall, as already mentioned less is more.
Obviously, this is a general guideline which cannot always be followed to the t. In multiple articles on this site, I am showing charts with 7 risk grades across them. This is violating the rule of 6, however, if there is no business logic in reducing certain category, sometimes you have to go along with it. On the charts below, you can decide for yourself, which is more readable.
... but if you have to break it - use a proper chart
If you have no way around having less than 6 series, do try to use a proper chart. There are ones who are designed to handle a lot of data. For example treemaps, sunbursts, stock charts and geo-maps (Excel 2016 supports both 3D and 2D geo maps as well as animations over them). Usually, you can get away with a lot more data this way, as long as you don’t g crazy with the labels.
3D charts for 2D data is a big NO
Whenever I see a chart, which is normally 2 dimensions, shown as 3D this is a red flag. It either means that someone is trying to lie with the chart or that they have no very little experience with presenting data. This might sound harsh, but the 3D charts are usually outright misleading. So unless you have a very good reason for doing so (your data is in 3 dimensions), you should NEVER use a 3D chart. You will actually get bonus points (with me at least) if you have 3D data and you manage to present it properly in a two-dimensional chart.
This extends, although to a much lesser extent, to other visuals such as shadows and bevel.
Data for the two charts below is the same. Obviously, this is an exaggerated example, but it still stands to prove 3D is misleading.
Don't play too much with the vertical axis scale
Similar to the topic above – this is a signal for foul play. It might be not intentional, you might just want to show data more clearly, but it doesn’t make it any less misleading. If you DO have to change the vertical axis, add a very clear (and visible) note that you have done so.
Data for the three charts below (as you can see from the labels) is 100% identical. The story which the charts tell – is not.
Colours are a matter of personal preference. I for one like pastel monochrome colours (although I acknowledge this article gives the exact opposite impression), because then my charts don’t look like a children’s colouring book. If your company does not have official colours for documents, you can adopt any approach, but I would recommend either using one of the pre-built pallets in your tool of choice or using the standard colour wheel techniques for using a theme (shown to the right).
Here are some other basic tips:
In case you have multiple charts try to use the same approach for all of them.
If in one chart you show sales as green and budget as red, use the same colour for them in other charts as well.
To complement your pastel colours, you can use single bring one to add a point of attention.
Be careful with transparency and overlapping series.
Try to develop your personal style over time and stick to it. In time people will recognise your work just by looking at it.
All that being said – don’t be a one-trick pony. Recognize that different cases call for a different approach, either depending on your audience or your message.
Basic colour-picking schemes.
Take your time with the details - it will pay off
As mentioned in the beginning – having good presentation is just as important as having good data. Here are some small details you can take care of:
Make sure the size of all your charts is the same or at least they have the same hight or width (depending on if they are displayed next to or above one another). It is easy to set up.
Try to keep the same text font and colour across your titles.
Make sure your labels overlap with your series nicely. Rotate them if needed.
Make sure there is no overlapping text or unreadable content.
Consistency is key for the small details. Try to keep the exact same style for gridlines, labels, titles and legends across all your charts.
Developing your own personal style for data presentation takes time and a lot of practice. But at the end, it is an immutable part of data analysis and one which is quite often used.